ポール・グレアム「プログラミングのFAQ」を翻訳しました。原題は Programming FAQ で、原文はココです。英語とプログラミングに強いみなさま、アドバイスをよろしくお願いいたします。なお翻訳にあたり、id:ttamo様、id:Shinnya様のアドバイスをいただいております。ありがとうございます!!

Programming FAQ

What editor do you use?


How can I learn to program?

Find a friend who knows how to program. Get them to set you up with a system where you can edit and run programs. Use whatever language they suggest for a beginner (probably it will be Python or Ruby). Then get the O'Reilly book and start working through it.

As you learn the mechanics of writing and running a program, start thinking about specific programs you want to write. That will motivate you to learn more.

Don't start with a problem that's too big. A good way to begin is to take an existing program and modify it to do something new.

Initially your programs will be ugly, but don't worry about that. Everyone's are. Just keep going, and they'll get better.

As you learn, you'll find it useful to look at programs other people have written. But you'll learn more from this once you've tried programming yourself.

Finally, find friends who like to write programs. They can answer your technical questions; you'll get new ideas from talking to them; and they'll be the audience for your first efforts.

Why do you advise plunging right into a programming project instead of carefully planning it first?

If you're trying to solve a simple, predefined problem like doing a depth-first search, thinking everything out beforehand doesn't hurt. But few real problems are like that. In real-world applications, you don't usually know at first precisely what problem you're trying to solve. So if you spend a lot of time planning in advance, what you'll end up with is a minutely detailed plan for solving the wrong problem.

With complex, ill-defined problems, you're better off writing a prototype as fast as you can, seeing what turns out to be wrong with it, and then changing your definition of the problem accordingly.

Often the reason programmers are pushed into planning is not that the problem requires it, but that project managers require it. Maybe programmers should give managers an explicit choice: do you want me to solve the problem in the way that will make you feel good, or the way that will yield the best solution?

Why do you keep going on about Lisp?

There are a number of topics I go on about: that small companies do things better than big ones; that cubicles suck; that you have to understand design to be a good hacker; that planning is overrated. Those don't seem so conspicuous, because many readers either have no prior opinion, or already agree.

キュービクルが最低だってことは誰にでもわかる。オフィスのレイアウトで、あるスタイルに固執する人もほとんどいない。だが言語では、みんな自分の知っている言語にすごく固執する。というのも、(a) 新しい言語を学ぶのはたいへんだし、(b) プログラミング言語はプログラムというものの考え方に影響するため、自分が慣れている言語よりも強力な言語は、想像さえ難しくなるからだ。
It doesn't cost anything to realize that cubicles suck. Few people have a vested interest in one style of office over another. But everyone has a vested interest in the languages they already know, because (a) it is a lot of work to learn a new language, and (b) programming languages dictate how you think about programs, so it is hard even to conceive of a language more powerful than whatever you're used to.

Dissing someone else's language is considered rude, I know. But rude is not the same as false. And any language designer has to face awkward questions like which languages are better, and why, just as proctologists have to do things that would be considered extremely rude if ordinary people did them.

Isn't object-oriented programming naturally suited to some problems?

Yes and no. A lot of what seem to be OO problems turn out not to be if you have random access to the concepts that together comprise object-orientedness.

If I were writing a CAD program or a simulation, for example, I'd probably use OO abstractions (though I'd probably end up creating my own OO model with macros instead of using whatever came with the language).

But if I were trying to solve the problem one reader sent to me as a canonical example of an OO problem, I wouldn't.


Suppose you have n serial ports, each of which may speak one of k protocols, and this must be configurable at run-time.

私だったら、単に n × k のクロージャ配列で表現するだろう。。
I'd just use an n-by-k array of closures to represent this.